Bitcoin mining economics have entered their most distressed phase since the April 2024 halving cut the block subsidy to 3.125 BTC. The weighted average ex-tax cash cost to produce one bitcoin among publicly listed miners reached $75,500 in Q2 2026, according to CoinShares, while BTC traded at $58...
"Listed Bitcoin miners that have pivoted to AI infrastructure are unlikely to return to mining, even if cryptocurrency prices continue to rise. For most operators, this pivot has become structurally irreversible." — Luke Nolan, CoinShares Research Analyst
Bitcoin mining economics have entered their most distressed phase since the April 2024 halving cut the block subsidy to 3.125 BTC. The weighted average ex-tax cash cost to produce one bitcoin among publicly listed miners reached $75,500 in Q2 2026, according to CoinShares, while BTC traded at $58,400 at quarter-end — placing the median listed miner below its cash breakeven line. Hashprice, the standard measure of mining revenue per unit of computing power, hit a five-year low of $27.70 per PH/s/day in June before recovering 43% to approximately $40 by late September as BTC climbed to $83,556.
The industry response has been a structural bifurcation. One group of operators — led by Core Scientific, Hut 8, and TeraWulf — has signed over $70 billion in AI and high-performance computing (HPC) lease contracts, backed in several cases by Google. Mining stock equities have outperformed BTC by 70 percentage points year-to-date, with a tracked basket up 56% while Bitcoin itself is down 17%. The other group, comprising legacy miners running older hardware, faces a binary outcome: upgrade to sub-15 J/TH ASICs at current rig prices, or exit. CoinShares estimates 15-20% of legacy rigs are now unprofitable even at October's improved hashprice levels.
Network hashrate, which peaked at 1,145 EH/s in October 2025, dropped 25% to 859 EH/s by January 2026 before stabilizing near 950 EH/s. Mining difficulty sits 18.3% below its peak — a drawdown that has lasted 322 days, the longest since 2012. The numbers describe an industry in forced restructuring: mining as a standalone business model is narrowing, while mining infrastructure as a dual-purpose compute platform is expanding.
CoinShares' Q2 2026 Bitcoin Mining Report laid out the arithmetic confronting the industry. The weighted average cash production cost for listed miners was approximately $75,500 per BTC. Bitcoin's quarter-end price of $58,400 meant most operators were producing at a loss on a cash-flow basis.
Individual company data confirmed the pattern. CleanSpark reported an all-in production cost of $135,691 per BTC, with a cash cost component of $71,995. Riot Platforms' all-in cost reached $113,499, with cash costs at $74,955. MARA Holdings posted Q2 revenue of $174.9 million, down 27% year-over-year from $238.5 million, with net losses widening to $611.3 million. CleanSpark's quarterly revenue fell 30.5% year-over-year to $138.0 million, with a net loss of $239.8 million.
The gap between all-in cost and cash cost reflects depreciation on hardware that was purchased at higher prices and is now generating less revenue per terahash. It also reflects SG&A costs that have not scaled down proportionally to revenue declines.
Hashprice — the dollar revenue a miner earns per petahash per second per day — is the industry's primary operating metric. It peaked at approximately $63/PH/s/day in July 2025, then declined steadily through Q4 2025 to $35-37 by November. By June 2026, it hit $27.70, a five-year low.
The driver was straightforward: record mining difficulty combined with a BTC price roughly 31% below its October 2025 all-time high and minimal transaction fee income. At $27.70, only operators running the latest generation of ASICs (sub-15 J/TH) at electricity costs below $0.06/kWh were generating positive cash margins.
September brought partial relief. Bitcoin climbed to approximately $83,556 by month-end, up 6.4% for the month. Hashprice recovered to an average of $39.38/PH/s/day in September, with a range between $37.49 and $41.32. The 43% increase from July lows restored profitability across all current-generation machines, though it left operators running older hardware — anything above 25 J/TH — in marginal territory.
The October 3 difficulty adjustment was expected to come in roughly flat, between -0.7% (Luxor estimate) and +0.5% (Mempool.space estimate), suggesting the September hashprice improvement should hold into early October absent a significant BTC price decline.
Bitcoin's network hashrate peaked at approximately 1,145 EH/s in October 2025. Between October and late January 2026, it dropped 25% to 859 EH/s as higher-cost operators shut down rigs and multiple companies redirected power capacity to AI workloads. By October 2026, hashrate had recovered to approximately 950 EH/s, but it remains well below the prior peak.
Mining difficulty, which adjusts every 2,016 blocks (approximately every two weeks), tracks hashrate with a lag. As of September 2026, difficulty sat at 132.76 T, which is 18.3% below the October 2025 peak. The drawdown has persisted for 322 days — the longest sustained decline since 2012.
The September adjustments were modestly positive: a 1.31% increase on September 5 and a 4.16% increase on September 19. These reflect the hashrate recovery from sub-900 EH/s levels earlier in the year but do not suggest a return to the prior peak is imminent. The geographic distribution of hashrate has also shifted: the United States accounts for 38-42% of global hashrate, Russia 16-17%, and China 14-15%, according to Hashrate Index Q2 2026 data.
Hardware efficiency has become the primary determinant of mining survival. The metric is joules per terahash (J/TH) — lower is better.
Bitmain's Antminer S23 Hyd, released in 2026, delivers 580 TH/s at 9.5 J/TH, making it the first mass-production SHA-256 miner to break the sub-10 J/TH barrier. The air-cooled S23 variant achieves 318 TH/s at 11 J/TH. The Antminer S21 XP, the most efficient air-cooled unit from the prior generation, runs at 270 TH/s and 13.5 J/TH.
For context, the industry standard as recently as Q1 2026 was 13-16 J/TH for current-generation machines. The 2020 generation operated at 30-40 J/TH. The original Antminer S9, dominant through 2016-2018, ran at 98 J/TH. Efficiency has improved approximately 85% over a decade.
CoinShares estimates that 15-20% of deployed rigs — primarily older-generation units above 25 J/TH — are now operating at a loss even at October's improved hashprice levels. Industrial miners running sub-15 J/TH hardware at electricity costs between $0.065 and $0.08/kWh report operating margins of 43-60%. The gap between these two cohorts defines the industry's current economic fault line.
The most significant structural change in Bitcoin mining in 2026 is the reallocation of power infrastructure from SHA-256 computation to AI and HPC workloads. As of mid-2026, signed AI/HPC contracts by Bitcoin mining companies exceeded $70 billion in total contract value, according to industry analysts.
The largest deals:
HIVE Digital estimated that 10 MW of Nvidia H100 GPU infrastructure generates revenue comparable to 100 MW of Bitcoin mining capacity — a 10:1 revenue density advantage that explains the economic logic of the transition.
The CoinShares Q2 2026 report concluded that this pivot is "structurally irreversible" for most operators: the capital commitments, lease terms, and GPU procurement contracts lock companies into HPC for 10-15 years. Even a sustained BTC price recovery above $100,000 would not alter the economic calculus for capacity already committed to AI tenants.
The revenue mix among listed mining companies is shifting rapidly. Two examples illustrate the pace:
Riot Platforms reported Q1 2026 revenue of $167.2 million, of which $111.9 million (67%) came from Bitcoin mining and $33.2 million (20%) from data center operations. The data center share is expected to grow as new HPC capacity comes online.
The consequence for Bitcoin's network is that power capacity previously available for hashrate growth is being permanently redirected. This is one factor behind the prolonged difficulty drawdown: capacity is not just going offline temporarily due to economics — it is being physically converted to serve different computational workloads.
The AI pivot has created a notable divergence between mining equity performance and the underlying cryptocurrency. Year-to-date through early October 2026, a tracked basket of the ten largest publicly traded mining stocks was up 56%, while Bitcoin was down approximately 17% — a 73-percentage-point spread.
Individual stock performance:
| Company | YTD Return (approx.) | |---------|---------------------| | TeraWulf | +85% | | Hut 8 | +67% | | Riot Platforms | +46% | | Basket Average | +56% | | Bitcoin | -17% |
The market is pricing mining companies primarily on their AI/HPC contract books rather than their Bitcoin production capacity. This represents a fundamental re-rating of the sector: mining stocks have decoupled from BTC as a leveraged play and are being valued as infrastructure-as-a-service operators with a residual cryptocurrency position.
The Bitcoin network's annualized electricity consumption ranges from 138 TWh to 204 TWh depending on methodology, according to estimates from the Cambridge Centre for Alternative Finance (CBECI) and the Digiconomist Bitcoin Energy Consumption Index. The CBECI places the figure at 170-180 TWh as of early 2026, equivalent to a continuous power draw of approximately 15.8-16.1 GW.
This is roughly equivalent to the electricity consumption of the Netherlands or Thailand and represents approximately 0.5% of global electricity production. An estimated 52.4% of the network's power consumption comes from sustainable energy sources, according to the Cambridge data.
The energy footprint has declined modestly from its October 2025 peak in proportion to the hashrate decline, but efficiency improvements in deployed hardware have partially offset the reduction in computing power. As older, less efficient rigs are replaced with sub-15 J/TH machines, the network can maintain or grow hashrate while consuming less total energy per terahash.
Bitcoin mining in Q4 2026 is no longer a single-product industry. The post-halving margin compression — block rewards cut in half while difficulty stayed elevated — forced a reckoning that the September hashprice recovery has only partially resolved. At $40/PH/s/day, current-generation hardware is profitable, but the long-term economic direction of the sector is being set by lease agreements, not by block rewards.
The $70 billion in signed AI/HPC contracts represents a permanent structural change in how mining infrastructure companies generate revenue. For the Bitcoin network, this means that hashrate growth will depend increasingly on BTC price appreciation and hardware efficiency gains rather than on new capacity buildout, because the new capacity is being allocated to AI tenants offering higher and more predictable revenue per megawatt.
The 322-day difficulty drawdown and the hashrate plateau at 950 EH/s are symptoms of this reallocation. The network is secure at current levels — 950 EH/s is by historical standards an enormous amount of computing power — but the growth trajectory that characterized the 2020-2025 period has been interrupted. Whether it resumes depends on whether BTC price levels can sustain mining economics that compete with AI workload economics on a per-megawatt basis. At current prices, they do not.